In computer science, there’s a concept known as “endianness”. It refers to the order in which data is stored: “big endian” means that we start with the most significant bit and move to the least; “little endian” is the other way around. For example, the decimal number 123, in 8-bits, can be represented as either 01111011 () in big endian, or 11011110 () in little. There are engineering reasons as to why one would choose one representation over the other.

Human, rather than binary encoded data often — at least from my anecdotal experience — follow the big endian model. For example, while this may be an artefact of the technology behind it, phone numbers follow a pattern like “country, area, number, extension”. Arguably, it’s much more logical as the most important part comes first, then tailing off. However, there are a number of inconstancies in some widely used formats; perhaps arising from the definition of “most important” differing from a purely quantitative scale.

Examples

Time and Date

In British English, we say the date as “day of the month, month, year”: today is 10^th January, 2012. This is little endian as the year is the category that has the most weight. Paradoxically, time (as in the time of day) is expressed in big endian: “hours, minutes, seconds, etc.”. Time of day and date are both forms of time — indeed, the date can be seen as the next order of magnitude, after time of day — so why the mix?

Presumably, this is because the time of day and date are not often expressed simultaneously, in every day conversation, and because we (or, at least, the British) have different priorities when referring to each. That is, it’s more important to know the day — something one tends to forget — over the year.

Of course, in other cultures, we see different models. Chinese, as we shall see in the next section, is consistently big endian: the (Western) date is expressed in the form “2012年1月10日”, the same order as time. This is also the format taken by the ISO, the International Standards Organisation, in ISO 8601.

Then there’s US English, to really muddle things, which expresses the date as “month, day of the month, year” (e.g., January 10^th, 2012). This ordering is grossly inconsistent — it’s neither big nor little endian — and my only assumption behind its inception is some kind of prescriptive or stylistic custom.

Postal Addresses

When we send a letter, in the West, the person we send it to is the most important, followed by their home, town, region and country. With the exception of postal codes, which are usually appended near the end of the address as a mechanical routing aid, this little endian format is ubiquitous.

In China, however, it’s written the other way around: the country comes first, reducing finally to the recipient’s name. At first glance, the big endian format appears to be a tad impersonal. Perhaps, but think how much easier it is for the postal service to route: they can look at the first line and know which district to send it to, then that district office can forward it to the appropriate city by reading the next line, continuing the process until we reach our destination. In a country the size of China, the big endian format could well speed up delivery time; which doesn’t seem so impersonal, after all!

Domain Names, URLs and E-Mail Addresses

Again, these examples are somewhat biased by their technological underpinnings, but bear with me!

The domain name system is little endian: the top level domain (.) comes at the right and we traverse downwards by writing leftwards. For example, in xoph.co, .xoph is a subdomain of .co, which sits below the root. Again, this is probably because the lower nodes in the domain tree are “more important”, in human terms, than the umbrella levels: You don’t care about .co — nothing sits there, anyway — but the .xoph subdomain is where things get interesting. This is also the way in which we, at least in English, read: e.g., the Amazon company, in the UK (amazon.co.uk), or the mail server at the Wikipedia organisation (mail.wikipedia.org), etc.

However, then it’s mixed into the big endian URL system, which starts with the domain name — which itself is in little endian but is the most significant part of the URL — followed by an increasingly variegated filesystem path. E-mail addresses, on the other hand, preserve the little endianness of the domain name only if the account name, which is largely arbitrary, maintains little endianness: e.g., j.doe@example.com is little endian, but doej@example.com is mixed.

Bookshelves

If you read from left-to-right, the front cover of a book is the left side and you progress through. When you’re done, you put your book on the shelf and, because the spine faces outwards, the front of the book is now on the right-hand-side. This effect is compounded with multiple volumes, which would be stacked left-to-right, but whose contents (from the perspective of someone looking at them in a bookshelf) are now mixed endian.

This effect isn’t limited to left-to-right writing systems. Right-to-left would have the same problem, just that everything is in the opposite direction. The only possible exception I can think of, when limited to horizontal writing systems, is modern Japanese: this is usually written left-to-right (like English), but starts on the far-right page (what a Westerner would call the last page) and works its way leftwards (towards the Western front page). Thus, when shelved, endianness is monotonic; but, of course, at the expense of a mixed endian page order!

The change in orientation, when shelving, means that we must always have mixed endianness somewhere, in this situation!